Built by the studio
that needed it.Not by a company entering a market.
Cutlist is made by HeckRaiser Productions, a solo photo and video studio. It started as an internal tool, because the person shooting the footage was spending more time hunting for clips than making them, and every tool the studio paid for did the same one thing: read the transcript.
HeckRaiser Productions. Solo, working, and the first user of everything here. Every default in the engine was set by someone who had to live with the output that weekend.
Encore comes out of the same studio and does the equivalent job for photographs. Same rules, same refusal to invent proof.
Statside, a multi-sport live-stats platform. It calls Cutlist over the API to cut league highlights, and it is the reason the API exists at all.
Three hours in.Eleven seconds out.
A broadcast runs long and most of it is nothing. Warmups, timeouts, an anthem, a stoppage, four minutes of a camera holding on an empty pitch. The parts worth posting are short, scattered, and completely obvious the second you see them - which is exactly why finding them by hand is intolerable. Scrub at 4x with a finger on the arrow keys, mark in, mark out, crop to vertical, burn captions, write the post, do it again twenty times. That is not editing. It is data entry with a playhead.
A transcript cannotsee a touchdown.
So the studio bought the automatic tools, and every one worked the same way: transcribe the audio, score the text, return timestamps. On a podcast that is defensible - the words are the content. On a rec-league stream with no commentary feed, on a concert, on a wedding, on a training session, the transcript is close to empty and a text score has nothing to rank. They did not fail loudly either. They returned confident clips of the wrong moments, which is worse than returning none.
The crowd
A stadium erupting is the loudest signal in the file and produces no words at all.
The scoreboard
The score, the clock, the down and distance are on screen, never in the transcript.
The ad break
Two minutes of commercial transcribes beautifully and is worth nothing.
The wide shot
A play develops across the frame. Text cannot tell you where it went.
See, listen,then read.
The fix was not a better text score. It was giving the thing eyes. Every candidate clip is sampled and looked at, the audio is measured for texture rather than volume, and the transcript joins as one input of three instead of being the whole case. A moment the crowd reacted to outranks one it did not. A moment whose frames turn out to be a commercial never ships.
See
Frames from inside the moment, including on-screen text a transcript never contains.
Listen
Spectral texture, not loudness. A crowd roar reads differently from a music sting.
Read
Word-level timings, which is what makes clip edges land on a sentence rather than mid-thought.
The long version, with what each sense actually measures, is on how it works.
Statside already knewwhen it happened.
Statside is a multi-sport live-stats platform: a scorekeeper logs every play as it happens, so the platform holds a timestamp for every moment worth posting before any model is asked anything. It sends those times to Cutlist and gets back vertical, captioned clips. That partnership is where the anchored path came from - nobody should pay a model to work out something they already wrote down.
Statside is the only partner named on this site because it is the only one there is. When there are others, they will appear here, and not before.
Five rules,no asterisks.
Original footage only
Your clip is your footage. No stock inserts, no synthetic filler, no generated b-roll - not because it is hard, but because a studio that shoots for a living does not ship pictures nobody took.
No invented proof
There are no testimonials on this site, no customer logos, no share-rate charts. Not because they would not help. Because we do not have them, and a made-up number is a lie with a decimal point in it.
The limits are written down
Where a feature has an edge, the edge is printed underneath it, and the things that are not built get a section of their own. A feature list made entirely of ticks tells you nothing.
Priced for the cut, not the pipeline
Every clip goes through a frontier model with the actual frames in hand. That costs materially more per minute than scoring a wall of text. Competing on price would mean shipping the cheaper pipeline, so we do not.
Hosted, and that is the whole product
You open a browser and clips come back. Nothing to install, nothing to keep running, no machine of yours doing the work.
The queue,without dates.
What is actually being worked on, in roughly this order. Nothing here is sold as if it shipped.
Shared seats and roles
Today it is one account and one set of projects. Editors, reviewers and per-project access are the largest thing missing.
Intro and outro cards
The brand kit already covers logo, colour, caption style and voice. Topping and tailing a clip does not exist yet.
More partner seams
The anchored path is general, not sports-specific. Anything that keeps a clock - event software, meeting tools, streaming overlays - can send times the same way.
Sharper reframing on crowds
Motion tracking holds a single subject well. Twenty players moving at once is where it is weakest, and that is where the work is.
No dates on any of it. A date given now would be a guess wearing a commitment. And generated b-roll is not on the list, this quarter or any other.
The trial is 160 credits and no card — about 34 minutes of footage through the whole engine, every feature on, no watermark on anything it hands back. Judge it on your own footage - that is the only test that counts.